A method and device for identifying knife switch status

Abstract

The present invention discloses a method and device for identifying the state of a knife switch. The identification method includes: establishing a standard image recognition model, enabling the standard image recognition model to identify the state of the knife switch in an image with the knife switch and determining the knife switch state recognition accuracy; photographing the same target knife switch at at least two shooting positions to obtain a group of actual images of the target knife switch; obtaining a position coordinate group, wherein the position coordinate group includes the position coordinates of the target knife switch relative to each shooting position; inputting the actual image group into the standard image recognition model and obtaining a recognition result group output by the standard image recognition model; outputting the final state of the target knife switch and the final state recognition accuracy based on the recognition result group and the position coordinate group. The present invention provides a method and device for identifying the state of a knife switch, which can improve the accuracy of knife switch state recognition and eliminate the need for manual on-site inspection of the knife switch state, thereby saving manpower and improving recognition efficiency.

Description

Technical Field

[0001] The present invention relates to the field of electric power technology, and in particular to a method and device for identifying a switch state. Background Art

[0002] When performing electrical operations on substation equipment, the operational quality of the equipment must be checked after each step to see if the equipment has reached the specified operating position. Only when the conditions are met can the next operation be carried out. For open-type equipment, since it operates in harsh outdoor environments for a long time, the equipment's operational reliability is not as good as that of indoor GIS equipment. Therefore, the inspection of its operational quality is more important, especially the inspection of the knife switch, which is more prominent. This is mainly due to its lack of arc extinguishing function and the tendency to generate heat defects when the switch is not closed properly. Currently, manual inspection is often used to determine whether the knife switch is closed. This manual inspection method is inefficient and has low recognition accuracy. Summary of the Invention

[0003] The present invention provides a method and device for identifying a knife switch state, which can improve the accuracy of knife switch state identification and eliminate the need for manual on-site inspection of the knife switch state, thereby saving manpower and improving identification efficiency.

[0004] According to one aspect of the present invention, a method for identifying a switch state is provided, the method comprising:

[0005] Establishing a standard image recognition model, enabling the standard image recognition model to recognize the state of a knife switch in an image with a knife switch and determining the knife switch state recognition accuracy, wherein the knife switch state includes closed or open;

[0006] photographing the same target knife switch at at least two photographing positions respectively to obtain a target knife switch actual image group, wherein the target knife switch actual image group includes the target knife switch actual image photographed at each of the photographing positions;

[0007] Acquire a position coordinate group, wherein the position coordinate group includes the position coordinates of the target knife switch relative to each of the shooting positions;

[0008] Inputting the actual image group into the standard image recognition model, and obtaining a recognition result group output by the standard image recognition model, wherein the recognition result group includes a recognition result unit of each actual image of the target knife switch, and the recognition result unit includes a target knife switch state and a recognition accuracy rate of the target knife switch state;

[0009] The final state of the target knife switch and the final state recognition accuracy are output according to the recognition result group and the position coordinate group.

[0010] Optionally, outputting the final state of the target knife switch and the final state recognition accuracy rate according to the recognition result group and the position coordinate group specifically includes:

[0011] Eliminating unqualified recognition result units in the recognition result group to obtain a state-consistent result group, wherein each target knife switch state in the state-consistent result group is the same, and the target knife switch state in the state-consistent result group is the final state of the target knife switch;

[0012] If the state consistent result group includes only one recognition result unit, the recognition accuracy rate of the target switch state in the recognition result unit is used as the final state recognition accuracy rate;

[0013] If the state-consistent result group includes more than one recognition result unit, the recognition result units are sorted in descending order according to the size of the target knife switch state recognition accuracy in the state-consistent result group; the shooting position corresponding to each of the sorted recognition result units and the connecting line of the target knife switch and the angle formed by the connecting line of the target shooting position and the target knife switch are determined according to the position coordinate group to obtain an angle group, wherein the target shooting position is the shooting position corresponding to the largest target knife switch state recognition accuracy in the state-consistent result group; the final state recognition accuracy is determined according to the largest target knife switch state recognition accuracy in the state-consistent result group and the angle group.

[0014] Optionally, determining the final state recognition accuracy rate based on the largest target switch state recognition accuracy rate and the angle group in the state consistent result group specifically includes:

[0015] The final state recognition accuracy is determined according to a preset condition; the preset condition is: X'=X(1+0.01a1 / 180°)…(1+0.01a n / 180°), where a1, ...a n are all the angles in the angle group, X is the maximum target switch state recognition accuracy in the state consistent result group, when X' is greater than 1, the final state recognition accuracy is 1, when X' is less than or equal to 1, X' is the final state recognition accuracy.

[0016] Optionally, the establishing of a standard image recognition model specifically includes:

[0017] Establish an image recognition model to be trained;

[0018] Inputting knife switch positive samples and knife switch negative samples into the to-be-trained image recognition model, and obtaining a training sample recognition result group output by the to-be-trained image recognition model, wherein the knife switch positive samples include sample images of the sample knife switch in the closed position, the knife switch negative samples include sample images of the sample knife switch in the open position, and the training sample recognition result group includes the sample knife switch state and the recognition accuracy of the sample knife switch state;

[0019] Adjusting the network parameters in the image recognition model to be trained according to the training sample recognition result group;

[0020] If the sample switch state recognition accuracy rate output by the adjusted image recognition model to be trained is greater than a preset value, the adjusted image recognition model to be trained is used as the standard image recognition model.

[0021] Optionally, the preset value is greater than or equal to 0.9.

[0022] Optionally, among the at least two shooting positions, an angle between a line connecting any two shooting positions and a reference point in the target knife switch is greater than or equal to a preset angle.

[0023] Optionally, the preset angle is greater than or equal to 30°.

[0024] Optionally, the knife switch includes a horizontal rotary knife switch, a vertical telescopic knife switch or a vertical scissor-type knife switch.

[0025] According to another aspect of the present invention, a device for identifying a switch state is provided, the device comprising a model building module, a photographing module, a position identification module, a recognition result acquisition module, and a final result output module;

[0026] The model building module is used to establish a standard image recognition model, so that the standard image recognition model can identify the status of a knife switch in an image with a knife switch and determine the recognition accuracy of the knife switch status, wherein the knife switch status includes closing or opening;

[0027] The shooting module is used to shoot the same target knife switch at at least two shooting positions respectively to obtain a target knife switch actual image group, wherein the target knife switch actual image group includes the target knife switch actual image shot at each of the shooting positions;

[0028] The position identification module is used to obtain a position coordinate group, wherein the position coordinate group includes the position coordinates of the target knife switch relative to each of the shooting positions;

[0029] The recognition result acquisition module is used to input the actual image group into the standard image recognition model and obtain the recognition result group output by the standard image recognition model, wherein the recognition result group includes a recognition result unit of each actual image of the target knife gate, and the recognition result unit includes the target knife gate state and the recognition accuracy rate of the target knife gate state;

[0030] The final result output module is used to output the final state of the target knife switch and the final state recognition accuracy rate according to the recognition result group and the position coordinate group.

[0031] Optionally, the recognition result acquisition module is specifically used to eliminate unqualified recognition result units in the recognition result group to obtain a state-consistent result group, wherein each target knife switch state in the state-consistent result group is the same, and the target knife switch state in the state-consistent result group is the final state of the target knife switch; if the state-consistent result group includes only one recognition result unit, the recognition accuracy rate of the target knife switch state in the recognition result unit is used as the final state recognition accuracy rate; if the state-consistent result group includes more than one recognition result unit, the recognition accuracy rate of the target knife switch state in the recognition result unit is used as the final state recognition accuracy rate. The recognition result units are sorted in descending order according to the size of the target knife switch state recognition accuracy; the angle formed by the line connecting the shooting position corresponding to each of the sorted recognition result units and the target knife switch and the line connecting the target shooting position and the target knife switch is determined according to the position coordinate group to obtain an angle group, wherein the target shooting position is the shooting position corresponding to the largest target knife switch state recognition accuracy in the state consistent result group; the final state recognition accuracy is determined according to the largest target knife switch state recognition accuracy in the state consistent result group and the angle group.

[0032] This embodiment provides a method for identifying a switch state. The method includes establishing a standard image recognition model, then photographing the same target switch at different shooting positions to obtain multiple actual images of the target switch taken at different shooting angles. All of the target switch actual images constitute a target switch actual image group. Next, the position coordinates of the target switch relative to each shooting position are obtained. All of the position coordinates constitute a position coordinate group. Each target switch actual image is input into the standard image recognition model, and the target switch state and target switch state recognition accuracy for each target switch actual image are obtained as output by the standard image recognition model. All of the target switch states and target switch state recognition accuracy rates constitute a recognition result group. Finally, the recognition result group and the position coordinate group are combined to determine the final state of the target switch and the final state recognition accuracy rate. It can be seen that the final state can be related to each target switch state and each target switch state recognition accuracy rate in the recognition result group, and the final state recognition accuracy rate can be related to each target switch state, each target switch state recognition accuracy rate, and the position coordinates of the target switch relative to the multiple shooting positions. This arrangement can improve the accuracy of target switch state recognition. In summary, the method for identifying the status of the knife switch provided in this embodiment can improve the accuracy of identifying the status of the knife switch, and does not require manual on-site checking of the status of the knife switch, thereby saving manpower and improving identification efficiency.

[0033] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0035] Figure 1 1 is a flow chart of a method for identifying a switch state according to an embodiment of the present invention;

[0036] Figure 2 1 is a schematic structural diagram of the relative positions of a target knife switch and different shooting positions provided according to an embodiment of the present invention;

[0037] Figure 3 1 is a schematic structural diagram of a device for identifying a switch state according to an embodiment of the present invention;

[0038] Figure 43 is a structural schematic diagram of the relative positions of a shooting module and a target knife switch provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0040] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0041] Figure 1 This is a flow chart of a method for identifying a switch state according to an embodiment of the present invention, referring to Figure 1 The identification method provided in this embodiment includes the following steps:

[0042] S110. Establish a standard image recognition model, so that the standard image recognition model recognizes the status of a knife switch in an image with a knife switch and determines the recognition accuracy of the knife switch status, wherein the knife switch status includes closing or opening.

[0043] Specifically, after inputting an image of a knife switch into a standard image recognition model, the standard image recognition model can identify the static and moving contacts of the knife switch in the image and determine the knife switch state based on the relative positions of the static and moving contacts. The knife switch state recognition accuracy can range from 0 to 1. The closer the knife switch state recognition accuracy is to 1, the higher the accuracy of the standard image recognition model in determining the knife switch state. When the knife switch state is inconsistent, the final state of the knife switch can be determined based on the knife switch state recognition accuracy. For example, two images of the same knife switch taken at different locations are input into the standard image recognition model. The standard image recognition model outputs one image as a closed switch with a knife switch state accuracy of 0.95, and the other image as an open switch with a knife switch state accuracy of 0.4. Since 0.95 is greater than 0.4, the accuracy of the knife switch state being closed is high. Therefore, based on these two results output by the standard image recognition model, the knife switch state is ultimately determined to be closed.

[0044] S120: Photograph the same target knife switch at at least two photographing positions respectively to obtain a target knife switch actual image group, wherein the target knife switch actual image group includes the target knife switch actual image photographed at each photographing position.

[0045] Specifically, a drone can be used to photograph the target switch at different latitudes and longitudes to obtain a set of actual images of the target switch. Alternatively, a drone and a fixed camera can be used to photograph the target switch to obtain a set of actual images of the target switch. By photographing the same target switch at different locations, multiple actual images of the target switch are obtained. The accuracy of the switch status determined based on these multiple actual images is higher, avoiding the randomness of the switch status determined based on a single actual image of the target switch.

[0046] Each target knife switch actual image includes the full image of the static contact and the full image of the movable contact of the target knife switch. If the target knife switch is three-phase, each target knife switch actual image can include the full image of the static contact and the full image of the movable contact of each of the three phases. Alternatively, if the target knife switch is three-phase, each phase knife switch must be photographed at at least two shooting positions, that is, each phase knife switch must have at least two target knife switch actual images. For example, if the target knife switch is three-phase, if the target knife switch actual image only includes one of the three phase knife switches, then at least 6 target knife switch actual images must be taken for the three-phase knife switches, and each phase knife switch must have at least two target knife switch actual images taken at different shooting positions.

[0047] S130: Acquire a position coordinate group, wherein the position coordinate group includes the position coordinates of the target knife switch relative to each shooting position.

[0048] Specifically, the position coordinates include the direction and the azimuth. For example, if the target knife switch is located 30° north of a shooting position, the position coordinates are (north by west, 30°). When shooting the target knife switch at each shooting position, the position coordinates of the target knife switch relative to the shooting position must be determined. The position coordinates can be determined by the shooting device that shoots the target knife switch.

[0049] S140. Input the actual image group into the standard image recognition model, and obtain the recognition result group output by the standard image recognition model, wherein the recognition result group includes a recognition result unit of each target knife switch actual image, and the recognition result unit includes the target knife switch state and the recognition accuracy of the target knife switch state.

[0050] Specifically, the actual image group includes multiple actual images of target knife switches. These multiple actual images of target knife switches are sequentially input into the standard image recognition model. The standard image recognition model outputs the corresponding target knife switch state and the target knife switch state recognition accuracy based on the target knife switch actual images. For example, the actual image group includes three actual images of target knife switches. The standard image recognition model can sequentially output the following: closed, 0.98; open, 0.5; closed, 0.8 based on these three actual images of target knife switches, where 0.98, 0.5, and 0.8 are the target knife switch state recognition accuracy rates for the three actual images of target knife switches, respectively.

[0051] S150. Output the final state of the target knife switch and the final state recognition accuracy rate according to the recognition result group and the position coordinate group.

[0052] Specifically, the recognition result group includes multiple target knife switch states, and the final state of the target knife switch can be determined based on the number of the same target knife switch states. For example, the recognition result group includes 5 recognition result units in total, 4 of the 5 recognition result units are open and 1 is closed, then the final state of the target knife switch can be determined as open based on the recognition result group.

[0053] The actual images of the target knife switch taken at different shooting positions may be different. The angle between the different shooting positions and the line connecting the target knife switch can be determined based on the position coordinate group, and the final state recognition accuracy of the target knife switch can be determined based on the angle and the recognition accuracy of each target knife switch state. The final state recognition accuracy can also reflect the degree of closing or opening of the target knife switch. For example, if the final state is closed and the final state recognition accuracy is 0.99, the final state recognition accuracy is relatively high. The staff can determine that the target knife switch is closed without going to the site to check the actual state of the target knife switch. The final state recognition accuracy of 0.99 can also indicate a high degree of closing. If the final state is closed and the final state recognition accuracy is 0.6, it means that the degree of closing is low, and it is possible that the closing is incomplete.

[0054] This embodiment provides a method for identifying a switch state. The method includes establishing a standard image recognition model, then photographing the same target switch at different shooting positions to obtain multiple actual images of the target switch taken at different shooting angles. All of the target switch actual images constitute a target switch actual image group. Next, the position coordinates of the target switch relative to each shooting position are obtained. All of the position coordinates constitute a position coordinate group. Each target switch actual image is input into the standard image recognition model, and the target switch state and target switch state recognition accuracy for each target switch actual image are obtained as output by the standard image recognition model. All of the target switch states and target switch state recognition accuracy rates constitute a recognition result group. Finally, the recognition result group and the position coordinate group are combined to determine the final state of the target switch and the final state recognition accuracy rate. It can be seen that the final state can be related to each target switch state and each target switch state recognition accuracy rate in the recognition result group, and the final state recognition accuracy rate can be related to each target switch state, each target switch state recognition accuracy rate, and the position coordinates of the target switch relative to the multiple shooting positions. This arrangement can improve the accuracy of target switch state recognition. In summary, the method for identifying the status of the knife switch provided in this embodiment can improve the accuracy of identifying the status of the knife switch, and does not require manual on-site checking of the status of the knife switch, thereby saving manpower and improving identification efficiency.

[0055] Optionally, outputting the final state and final state recognition accuracy of the target knife switch based on the recognition result group and the position coordinate group specifically includes: eliminating unqualified recognition result units in the recognition result group to obtain a state-consistent result group, wherein each target knife switch state in the state-consistent result group is the same, and the target knife switch state in the state-consistent result group is the final state of the target knife switch; if the state-consistent result group includes only one recognition result unit, the target knife switch state recognition accuracy in the recognition result unit is used as the final state recognition accuracy; if the state-consistent result group includes more than one recognition result unit, the recognition result units are sorted in descending order according to the size of the target knife switch state recognition accuracy in the state-consistent result group; determining the angle formed by the connecting line of the shooting position corresponding to each sorted recognition result unit and the target knife switch and the connecting line of the target shooting position and the target knife switch according to the position coordinate group to obtain an angle group, wherein the target shooting position is the shooting position corresponding to the largest target knife switch state recognition accuracy in the state-consistent result group; determining the final state recognition accuracy according to the largest target knife switch state recognition accuracy and the angle group in the state-consistent result group.

[0056] Specifically, in the recognition result group, if the number of recognition result units whose target switch state is closed is less than the number of recognition result units whose target switch state is open, the recognition result unit whose target switch state is closed is an unqualified recognition result unit. If the number of recognition result units whose target switch state is open is less than the number of recognition result units whose target switch state is closed, the recognition result unit whose target switch state is open is an unqualified recognition result unit. If all target switch states in the recognition result group are the same, there is no unqualified recognition result unit in the recognition result group. If the number of recognition result units with the target knife switch state being closed is equal to the number of recognition result units with the target knife switch state being open, then the sum of the recognition accuracy rates of all target knife switch states under closed condition is compared with the sum of the recognition accuracy rates of all target knife switch states under open condition, and the smaller one is regarded as an unqualified recognition result unit. For example, the recognition result group includes 6 recognition result units, namely: closed condition, 0.98; open condition, 0.5; closed condition, 0.8; closed condition, 0.85, open condition, 0.6; open condition , 0.7; It can be seen that the number of closing switches is equal to the number of opening switches. The sum of the recognition accuracy rates of all target knife switch states under closing is 2.63, and the sum of the recognition accuracy rates of all target knife switch states under opening is 1.8. Since 2.63 is greater than 1.8, the recognition result unit with the target knife switch state as opening is regarded as an unqualified recognition result unit. After eliminating the unqualified recognition result units in the recognition result group, the obtained state consistent result group is: closing, 0.98; closing, 0.8; closing, 0.85.

[0057] After obtaining a state-consistent result group including multiple recognition result units, the recognition result units are sorted according to the target switch state recognition accuracy. For example, if the state-consistent result group obtained after eliminating unqualified recognition result units is: closed, 0.98; closed, 0.8; closed, 0.85; then after sorting, the order of the three recognition result units is: 1. Closed, 0.98; 2. Closed, 0.85; 3. Closed, 0.8. The shooting position corresponding to the target switch state recognition accuracy of 0.98 is the target shooting position, that is, the shooting position ranked first is the target shooting position. After sorting, determine the angle formed by the target shooting position and the line connecting each shooting position and the target knife switch. Specifically, determine the angle formed by the target shooting position and the line connecting the shooting position ranked first and the target knife switch according to the position coordinate group. The angle is 0°. Determine the angle formed by the target shooting position and the line connecting the shooting position ranked second and the target knife switch. Determine the angle formed by the target shooting position and the line connecting the shooting position ranked third and the target knife switch to obtain an angle group.

[0058] Since the maximum target knife switch state recognition accuracy is higher, the accuracy of determining the final state recognition accuracy based on the maximum target knife switch state recognition accuracy and angle group in the state consistent result group is also higher, which can better reflect the actual state of the target knife switch.

[0059] Optionally, determining the final state recognition accuracy based on the maximum target switch state recognition accuracy and the angle group in the state consistent result group specifically includes: determining the final state recognition accuracy based on a preset condition; the preset condition is: X'=X(1+0.01a1 / 180°)…(1+0.01a n / 180°), where a1, ...a n are all the angles in the angle group, X is the maximum target switch state recognition accuracy in the state consistent result group, when X' is greater than 1, the final state recognition accuracy is 1, when X' is less than or equal to 1, X' is the final state recognition accuracy.

[0060] For example, after sorting, if the order of the recognition result units in the state consistent result group is: No. 1: closed, 0.98; No. 2: closed, 0.8. And the angle between the shooting position of the recognition result unit No. 1 (the shooting position corresponding to the recognition result unit No. 1 is the target shooting position) and the shooting position of the recognition result unit No. 2 and the target knife switch (hereinafter, for convenience, the angle is referred to as the angle between No. i and No. i+1) is 60°, then X' = 0.98(1+0.01*0° / 180°)(1+0.01*60° / 180°) = 0.983, and the final state recognition accuracy is 0.983.

[0061] After sorting, if the order of the recognition result units in the state consistent result group is: No. 1: closed, 0.98; No. 2: closed, 0.94; No. 3: closed, 0.9; No. 4: closed, 0.8; it can be seen that the shooting position corresponding to the recognition result unit No. 1 is the target shooting position, and according to the position coordinate group, it is determined that the angle between No. 1 and No. 2 is 60°, the angle between No. 1 and No. 3 is 40°, and the angle between No. 1 and No. 4 is 90°, then X'=0.98(1+0.01*0° / 180°)(1+0.01*60° / 180°)(1+0.01*40° / 180°)(1+0.01*90° / 180°)=0.99, and the final state recognition accuracy is 0.99.

[0062] In summary, the final state recognition accuracy provided by this embodiment is related to the shooting position and the maximum target knife switch state recognition accuracy. Such a setting can improve the accuracy of the final state recognition accuracy.

[0063] It should be noted that when obtaining the final state recognition accuracy, X' can be calculated in sequence according to the order of arrangement in the state consistent result group after sorting, and the calculation can be stopped when X' is greater than or equal to 1. For example, after sorting, if the order of the recognition result units in the state consistent result group is: No. 1: closed, 0.99; No. 2: closed, 0.98; No. 3: closed, 0.97; No. 4: closed, 0.96; and the angle between the target shooting position and No. 2 is 180°, the angle between the target shooting position and No. 3 is 150°, and the angle between the target shooting position and No. 4 is 100°. Then, according to the calculation results of numbers 1 and 2, 0.99(1+0.01*180° / 180°)=0.9999 is obtained. According to the calculation results of numbers 1 and 2 and the angle between the target shooting position and number 3, X'=0.9999(1+0.01*150° / 180°)=1.008 is calculated. Then the calculation is stopped and the final state recognition accuracy is output as 1.

[0064] Optionally, establishing a standard image recognition model specifically includes: establishing an image recognition model to be trained; inputting knife switch positive samples and knife switch negative samples into the image recognition model to be trained, and obtaining a training sample recognition result group output by the image recognition model to be trained, wherein the knife switch positive samples include sample images of the sample knife switch in the closed position, the knife switch negative samples include sample images of the sample knife switch in the open position, and the training sample recognition result group includes the sample knife switch state and the sample knife switch state recognition accuracy; adjusting the network parameters in the image recognition model to be trained according to the training sample recognition result group; if the sample knife switch state recognition accuracy output by the adjusted image recognition model to be trained is greater than a preset value, the adjusted image recognition model to be trained is used as the standard image recognition model.

[0065] Specifically, when the knife switch is in the closed position, it means that the knife switch state is closed, and when the knife switch is in the open position, it means that the knife switch state is open.

[0066] Optional, preset value greater than or equal to 0.9.

[0067] Specifically, when the preset value is greater than or equal to 0.9, it indicates that the adjusted image recognition model to be trained can accurately identify the relative positions of the static contact and the moving contact in the image with the knife switch, and thus has a high recognition accuracy rate for the knife switch state.

[0068] Optionally, among the at least two shooting positions, an angle between a line connecting any two shooting positions and a reference point in the target knife switch is greater than or equal to a preset angle.

[0069] Specifically, Figure 2 is a schematic structural diagram of the relative positions of a target knife switch and different shooting positions according to an embodiment of the present invention, with reference to Figure 2 , select a point in the target switch as a reference point, Figure 2 In the example, the reference point is denoted as b. Different shooting positions are connected to the reference point. The lines connecting each two shooting positions and the reference point have an angle. Figure 2 In the example, the angle between shooting positions 1 and 2 and the reference point of the target knife switch is denoted as c. Setting the angle between any two shooting positions and the reference point of the target knife switch to be greater than or equal to a preset angle can improve the problem of the actual images of the target knife switch taken at different shooting positions being approximately the same due to the proximity of the shooting positions, thereby improving the problem of low recognition rate of the target knife switch status caused by the proximity of the shooting positions.

[0070] Optionally, the preset angle is greater than or equal to 30°.

[0071] Specifically, setting the preset angle to be greater than or equal to 30° can improve the problem of low recognition rate of the target knife switch status caused by close shooting positions.

[0072] Optionally, the knife gate includes a horizontal rotary knife gate, a vertical telescopic knife gate or a vertical scissor-type knife gate.

[0073] Specifically, the horizontal rotary knife switch, vertical telescopic knife switch or vertical scissor knife switch in the closed or open state can be identified using the identification method provided in this embodiment.

[0074] Figure 3 Schematic diagram of a switch state identification device according to an embodiment of the present invention, Figure 3 The recognition device includes: a model building module 110, a shooting module 120, a position recognition module 130, a recognition result acquisition module 140 and a final result output module 150; the model building module 110 is used to establish a standard image recognition model, so that the standard image recognition model recognizes the state of the knife switch in the image with the knife switch and determines the recognition accuracy of the knife switch state, wherein the knife switch state includes closing or opening; the shooting module 120 is used to shoot the same target knife switch at at least two shooting positions respectively to obtain a target knife switch actual image group, wherein the target knife switch actual image group includes the target knife switch actual image shot at each shooting position image; the position recognition module 130 is used to obtain a position coordinate group, wherein the position coordinate group includes the position coordinates of the target knife switch relative to each shooting position; the recognition result acquisition module 140 is used to input the actual image group into the standard image recognition model, and obtain the recognition result group output by the standard image recognition model, wherein the recognition result group includes a recognition result unit of each target knife switch actual image, and the recognition result unit includes the target knife switch state and the target knife switch state recognition accuracy; the final result output module 150 is used to output the final state of the target knife switch and the final state recognition accuracy according to the recognition result group and the position coordinate group.

[0075] Specifically, Figure 4Schematic diagram of the relative position of a shooting module and a target knife switch according to an embodiment of the present invention, Figure 4 The shooting module provided in this embodiment may include a fixed camera, a drone, a patrol robot and a spherical camera. The shooting module is used to shoot the target knife switch 100 to obtain an actual image of the target knife switch.

[0076] Optionally, the recognition result acquisition module is specifically used to eliminate unqualified recognition result units in the recognition result group to obtain a state-consistent result group, wherein each target knife switch state in the state-consistent result group is the same, and the target knife switch state in the state-consistent result group is the final state of the target knife switch; if the state-consistent result group includes only one recognition result unit, the target knife switch state recognition accuracy in the recognition result unit is used as the final state recognition accuracy; if the state-consistent result group includes more than one recognition result unit, the recognition result units are sorted in descending order according to the size of the target knife switch state recognition accuracy in the state-consistent result group; according to the position coordinate group, the angle formed by the line connecting the shooting position corresponding to each sorted recognition result unit and the target knife switch and the line connecting the target shooting position and the target knife switch is determined to obtain an angle group, wherein the target shooting position is the shooting position corresponding to the largest target knife switch state recognition accuracy in the state-consistent result group; the final state recognition accuracy is determined according to the largest target knife switch state recognition accuracy in the state-consistent result group and the angle group. The device for identifying the status of the knife switch provided in an embodiment of the present invention has corresponding beneficial effects as does the method for identifying the status of the knife switch provided in any embodiment of the present invention. For technical details not detailed in this embodiment, please refer to the method for identifying the status of the knife switch provided in any embodiment of the present invention.

[0077] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present invention can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved. This is not limited herein.

[0078] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A method for identifying the status of a knife switch, characterized in that: include: Establishing a standard image recognition model, enabling the standard image recognition model to recognize the state of a knife switch in an image with a knife switch and determining the knife switch state recognition accuracy, wherein the knife switch state includes closed or open; photographing the same target knife switch at at least two photographing positions respectively to obtain a target knife switch actual image group, wherein the target knife switch actual image group includes the target knife switch actual image photographed at each of the photographing positions; Acquire a position coordinate group, wherein the position coordinate group includes the position coordinates of the target knife switch relative to each of the shooting positions; Inputting the actual image group into the standard image recognition model, and obtaining a recognition result group output by the standard image recognition model, wherein the recognition result group includes a recognition result unit of each actual image of the target knife switch, and the recognition result unit includes a target knife switch state and a recognition accuracy rate of the target knife switch state; Outputting the final state of the target knife switch and the final state recognition accuracy rate according to the recognition result group and the position coordinate group; Determining the final state recognition accuracy based on the largest target switch state recognition accuracy and angle group in the state consistent result group specifically includes: The final state recognition accuracy is determined according to a preset condition; the preset condition is: X'=X(1+0.01a1 / 180°)…(1+0.01a n / 180°), where a1, ...a n are all the angles in the angle group, X is the maximum target switch state recognition accuracy in the state consistent result group, when X' is greater than 1, the final state recognition accuracy is 1, when X' is less than or equal to 1, X' is the final state recognition accuracy.

2. The identification method according to claim 1, characterized in that Outputting the final state of the target knife switch and the final state recognition accuracy rate according to the recognition result group and the position coordinate group specifically includes: Eliminating unqualified recognition result units in the recognition result group to obtain a state-consistent result group, wherein each target knife switch state in the state-consistent result group is the same, and the target knife switch state in the state-consistent result group is the final state of the target knife switch; If the state consistent result group includes only one recognition result unit, the recognition accuracy rate of the target switch state in the recognition result unit is used as the final state recognition accuracy rate; If the state-consistent result group includes more than one recognition result unit, the recognition result units are sorted in descending order according to the size of the target knife switch state recognition accuracy in the state-consistent result group; the shooting position corresponding to each of the sorted recognition result units and the connecting line of the target knife switch and the angle formed by the connecting line of the target shooting position and the target knife switch are determined according to the position coordinate group to obtain an angle group, wherein the target shooting position is the shooting position corresponding to the largest target knife switch state recognition accuracy in the state-consistent result group; the final state recognition accuracy is determined according to the largest target knife switch state recognition accuracy in the state-consistent result group and the angle group.

3. The identification method according to claim 1, characterized in that The establishment of the standard image recognition model specifically includes: Establish an image recognition model to be trained; Inputting knife switch positive samples and knife switch negative samples into the to-be-trained image recognition model, and obtaining a training sample recognition result group output by the to-be-trained image recognition model, wherein the knife switch positive samples include sample images of the sample knife switch in the closed position, the knife switch negative samples include sample images of the sample knife switch in the open position, and the training sample recognition result group includes the sample knife switch state and the recognition accuracy of the sample knife switch state; Adjusting the network parameters in the image recognition model to be trained according to the training sample recognition result group; If the sample switch state recognition accuracy rate output by the adjusted image recognition model to be trained is greater than a preset value, the adjusted image recognition model to be trained is used as the standard image recognition model.

4. The identification method according to claim 3, characterized in that The preset value is greater than or equal to 0.

9.

5. The identification method according to claim 1, characterized in that Among the at least two shooting positions, an angle between a line connecting any two shooting positions and a reference point in the target knife switch is greater than or equal to a preset angle.

6. The identification method according to claim 5, characterized in that The preset angle is greater than or equal to 30°.

7. The identification method according to claim 1, characterized in that: The knife switch includes a horizontal rotary knife switch, a vertical telescopic knife switch or a vertical scissor-type knife switch.

8. A device for identifying the status of a knife switch, characterized in that: The identification method according to any one of claims 1 to 7 is used to identify the status of the knife switch, and the device includes a model building module, a shooting module, a position identification module, an identification result acquisition module and a final result output module; The model building module is used to establish a standard image recognition model, so that the standard image recognition model can identify the status of a knife switch in an image with a knife switch and determine the recognition accuracy of the knife switch status, wherein the knife switch status includes closing or opening; The shooting module is used to shoot the same target knife switch at at least two shooting positions respectively to obtain a target knife switch actual image group, wherein the target knife switch actual image group includes the target knife switch actual image shot at each of the shooting positions; The position identification module is used to obtain a position coordinate group, wherein the position coordinate group includes the position coordinates of the target knife switch relative to each of the shooting positions; The recognition result acquisition module is used to input the actual image group into the standard image recognition model and obtain the recognition result group output by the standard image recognition model, wherein the recognition result group includes a recognition result unit of each actual image of the target knife gate, and the recognition result unit includes the target knife gate state and the recognition accuracy rate of the target knife gate state; The final result output module is used to output the final state of the target knife switch and the final state recognition accuracy rate according to the recognition result group and the position coordinate group.

9. The identification device according to claim 8, characterized in that The recognition result acquisition module is specifically used to eliminate unqualified recognition result units in the recognition result group to obtain a state-consistent result group, wherein each target knife switch state in the state-consistent result group is the same, and the target knife switch state in the state-consistent result group is the final state of the target knife switch; if the state-consistent result group includes only one recognition result unit, the target knife switch state recognition accuracy in the recognition result unit is used as the final state recognition accuracy; if the state-consistent result group includes more than one recognition result unit, the recognition result units are sorted in descending order according to the size of the target knife switch state recognition accuracy in the state-consistent result group; according to the position coordinate group, the angle formed by the line connecting the shooting position corresponding to each sorted recognition result unit and the target knife switch and the line connecting the target shooting position and the target knife switch is determined to obtain an angle group, wherein the target shooting position is the shooting position corresponding to the largest target knife switch state recognition accuracy in the state-consistent result group; the final state recognition accuracy is determined according to the largest target knife switch state recognition accuracy in the state-consistent result group and the angle group.

Images